Computer Vision News - August 2020

39 Nicola Rieke 7 Best of MIDL 2020 that we all have to work on this together. This means we have to communicate. We have to exchange. When I talk to researchers, I want to learn from them. I want to understand what their issues are, what they are working on, their difficulties. When I talk to startups, I am curious about their ideas, and maybe I can help them with their solutions. When I talk to hospitals, it’s completely different. I talk to doctors. They say, “Well I heard about AI. What can we do for that? How can we enable that?” There are a lot of different questions and different levels of expertise and communities. By now our readers understand that you have at least one foot in healthcare, maybe even two. Can you tell us why NVIDIA is involved specifically in healthcare? What are the specific needs for that? NVIDIA actually has a dedicated healthcare global team. Healthcare is very different from automotive or gaming. The customers in the end are the patients. This has very different requirements than if you were to develop an app for the phone, for example, or a self-driving car. To fit this need, you really have to understand the problem. I think it’s very particular. You need domain expertise in this area to really understand the problem. Why the hardware should be different? Or is it related to software? NVIDIA in the past was really known for its hardware, especially for its PC gaming platform. We have developed quite a lot since then. It’s a computing platform company now. To really enable healthcare for the patient, you have to enable the full stack, starting from hardware. So for example, on the lower level the hardware, then CUDA as the software platform. Then a lot of acceleration libraries. If you work in deep learning, you probably use cuDNN on the back end. It can go higher and higher, until libraries. For example, all the software tools that we’ve developed for healthcare are within the umbrella of our CLARA application framework. We also work actively with the research community, so for example with MONAI , the Medical Open Network for AI that you also reviewed in your July magazine.